DEEP LEARNING WITH MULTIBAND SYNTHESIS FROM LANDSAT-8 SATELLITE IMAGERY USING MACHINE LEARNING

Abstract

This research article proposes a novel deep learning representation and segmentation approach for moderate resolution remote sensing image analysis. A data extraction approach using deep hierarchical understanding for remote sensing image is adopted as a test bed for further increase in spatial resolution imagery. The idea is the fact that we can adopt a quick scanning image segmentation in a deep learning feature representation framework using a deep learning technique to produce reasonable sized clusters in segmented regions until it forms a super-object. Our contribution is to implement an effective procedure for multi-scale image analysis to address the issue of measuring uncertainty in practice.

We then propose to test our method on two high resolution remote sensing image datasets that will output results in the form of multi-layered scenes that attest the efficiency and reliability of our proposed system.